--- license: cc-by-nc-4.0 task_categories: - tabular-classification - time-series-forecasting tags: - cybersecurity - identity-security - account-takeover - mfa-bypass - ueba - zero-trust - apt - synthetic-data - lateral-movement - golden-ticket pretty_name: CYB006 — Synthetic Login Activity Dataset (Sample) size_categories: - 1K 🤖 **Trained baseline available:** > [**xpertsystems/cyb006-baseline-classifier**](https://huggingface.co/xpertsystems/cyb006-baseline-classifier) > — XGBoost + PyTorch MLP for **3-class user-risk-tier classification** > (insider-threat scoring use case), stratified split, multi-seed evaluation > (ROC-AUC 0.812 ± 0.048). **Includes a structural-leakage diagnostic on > the threat-actor detection task** that buyers planning ATO / threat-actor > ML work should read first. | File | Rows (sample) | Rows (full) | Description | |----------------------------|---------------|---------------|----------------------------------------------| | `identity_topology.csv` | ~150 | ~3,200 | Identity domain registry | | `user_risk_summary.csv` | ~200 | ~6,500 | Per-user risk aggregates | | `login_sessions.csv` | ~5,000 | ~377,000 | Per-session login records (primary file) | | `auth_events.csv` | ~31,900 | ~750,000 | Discrete authentication event log | ## Dataset Summary CYB006 simulates enterprise login activity as a **6-phase session state machine** across diverse identity infrastructures, with: - **4 threat-actor capability tiers**: script_kiddie, opportunistic, advanced_persistent_threat (APT), nation_state — with per-tier credential attack patterns, MFA bypass propensity, lateral hop distributions, and Golden Ticket / Pass-the-Hash abuse rates - **8 identity domain types**: on-premises AD, Azure AD, Okta, hybrid_joined, SAML federated, zero_trust_ztna, PAW (privileged access workstation), SaaS application portal — each with distinct detection_strength and resilience scores - **MFA challenge methods**: disabled, SMS, TOTP, push notification, phishing-resistant FIDO2 — with per-method bypass propensity calibration - **6 session lifecycle phases**: pre_auth_probe, credential_submission, mfa_challenge, session_active, lateral_traversal, session_termination - **Geo-velocity modeling** with impossible travel detection via Haversine distance and per-user expected geolocation baselines - **UEBA scoring** with calibrated false-positive rates - **Conditional Access (CA) policy enforcement** modeling — ZTNA block strength tunable per architecture ## Trained Baseline Available A working baseline classifier trained on this sample is published at **[xpertsystems/cyb006-baseline-classifier](https://huggingface.co/xpertsystems/cyb006-baseline-classifier)**. | Component | Detail | |---|---| | Primary task | **3-class user_risk_tier classification** (insider-threat scoring) | | Diagnostic | Audit of threat-actor detection on this sample (see `leakage_diagnostic.json`) | | Models | XGBoost (`model_xgb.json`) + PyTorch MLP (`model_mlp.safetensors`) | | Features | 34 per-user features (aggregates + non-leaky session aggregates + engineered) | | Split | **Stratified by user_risk_tier** — user-level task, n=200 | | Validation | Single seed + multi-seed aggregate across 10 seeds | | Demo | `inference_example.ipynb` — end-to-end copy-paste | | Headline metrics | XGBoost: accuracy 0.700 ± 0.082, macro ROC-AUC 0.812 ± 0.048 (multi-seed) | **Important diagnostic finding for buyers planning threat-actor detection work:** the model card documents that this sample's threat-actor-vs-legitimate session populations have **non-overlapping anomaly score distributions** across at least six feature groups (velocity, timestamp, credential attempt count, login outcome, geo country, device trust). As a result, a plain XGBoost achieves 100% test accuracy on threat-actor binary detection that does not reflect real-world detection difficulty. The baseline model targets `user_risk_tier` instead, which is a legitimate ML task on the sample. See the model card's [Leakage diagnostic](https://huggingface.co/xpertsystems/cyb006-baseline-classifier#leakage-diagnostic) section for the full audit and recommendations. ## Calibrated Benchmark Targets The full product is calibrated to **12 benchmark validation tests** drawn from authoritative identity security sources (Microsoft Digital Defense Report, Okta Customer Identity Trends, Verizon DBIR, CISA Joint Advisories, Mandiant M-Trends, MITRE ATT&CK Evaluations, Gartner IAM Hype Cycle, KuppingerCole Leadership Compass). **Benchmark categories** (calibrated in both sample and full product): 1. **Credential attack velocity** — brute force (~50 RPS), password spray (<1 RPS) 2. **Account takeover rate by tier** — graduated by attacker capability 3. **MFA bypass rate** — FIDO2 ≤1%, push/SMS variable 4. **Impossible travel rate** — 7-12% of sessions 5. **Lateral movement depth** — capped per tier (script_kiddie ≤1.2 → nation_state ≤14) 6. **Privilege escalation rate** — conditional on lateral movement 7. **MFA fatigue burst timing** — Poisson λ=7 burst pattern 8. **UEBA false positive rate** — calibrated to 10-14% range 9. **Golden Ticket / Pass-the-Hash detection gap** — stealth modeling 10. **Session duration anomaly separation** — KL divergence proxy 11. **Conditional Access block rate** — ZTNA ≥88% for untrusted 12. **Kill-chain completion rate** — phase-to-phase progression Sample benchmark results: | Test | Description | Verdict | |------|-------------|---------| | T01 | Credential Attack Velocity | ✓ PASS | | T02 | Account Takeover Rate by Tier | ✓ PASS | | T03 | MFA Bypass Rate (FIDO2) | ✓ PASS | | T04 | Impossible Travel Rate | ✓ PASS | | T05 | Lateral Movement Depth by Tier | ✓ PASS | | T06 | Privilege Escalation Rate | ✓ PASS | | T07 | MFA Fatigue Burst Detection | ✓ PASS | | T08 | UEBA False Positive Rate | ✓ PASS | | T09 | Golden Ticket / PtH Detection Gap | ✓ PASS | | T10 | Session Duration Anomaly Separation | ✓ PASS | | T11 | Conditional Access Block Rate (ZTNA) | ✓ PASS | | T12 | Kill-Chain Completion Rate | ✓ PASS | *Note: some benchmarks (e.g. nation-state account takeover rates, Golden Ticket detection) require larger sample sizes to converge tightly because they're conditional on small attacker-tier subsets (nation_state ≈ 2% of all sessions, APT ≈ 3%). The full product demonstrates all 12 benchmarks with strong statistical power.* ## Schema Highlights ### `login_sessions.csv` (primary file) | Column | Type | Description | |---------------------------------|---------|----------------------------------------------| | session_id | string | Unique session identifier | | user_id | string | User identifier (FK to user_risk_summary) | | session_timestamp_utc | string | ISO timestamp | | session_phase | string | 1 of 6 phases | | login_outcome | string | success / failed / mfa_required / blocked | | source_ip_hash | string | SHA-256 pseudonymised source IP | | geo_country_code | string | ISO 3166 country code | | geo_city_hash | string | Hashed city locator | | device_id_hash | string | Hashed device fingerprint | | device_trust_level | string | unknown / known / managed / compliant | | authentication_method | string | password / sso / certificate / api_key | | mfa_challenge_type | string | disabled / sms / totp / push / fido2 | | mfa_response_latency_ms | int | MFA response latency | | credential_attempt_count | int | Attempts before success | | session_duration_seconds | int | Session length | | target_application_id | string | Application accessed | | privilege_level_accessed | string | standard / power_user / admin / domain_admin | | user_risk_tier | string | low / medium / high / critical | | threat_actor_capability_tier | string | script_kiddie / opportunistic / apt / nation_state (target) | | geo_anomaly_score | float | Geographic anomaly score (0–1) | | velocity_anomaly_score | float | Login velocity anomaly score (0–1) | | impossible_travel_flag | int | Boolean — impossible travel detected | ### `user_risk_summary.csv` (per-user aggregates) | Column | Type | Description | |---------------------------------|---------|----------------------------------------------| | user_id | string | User identifier | | user_risk_tier | string | Risk tier classification target | | total_login_attempts | int | Total login attempts in window | | successful_logins | int | Successful logins | | failed_logins | int | Failed logins | | mfa_failures | int | MFA challenge failures | | impossible_travel_events | int | Count of impossible travel detections | | lateral_hop_count | int | Total lateral movement hops | | privilege_escalations | int | Privilege escalation count | | account_lockout_count | int | Account lockout events | | geo_dispersion_score | float | Geographic dispersion (0–1) | | login_velocity_score | float | Velocity anomaly (0–1) | | session_anomaly_rate | float | Fraction of anomalous sessions | | ueba_alert_count | int | UEBA alerts raised | | threat_actor_flag | int | Boolean — threat actor | | account_takeover_flag | int | Boolean — account takeover detected | | overall_identity_risk_score | float | Composite identity risk (0–1) | | insider_threat_indicator_score | float | Insider threat composite (0–1) | See `auth_events.csv` and `identity_topology.csv` for the event log and identity domain schemas respectively. ## Suggested Use Cases - Training **insider threat scoring** models — [worked example available](https://huggingface.co/xpertsystems/cyb006-baseline-classifier) - **Account takeover (ATO) detection** model development (see leakage diagnostic in the baseline model card before training) - **Threat-actor tier classification** — 4-class with realistic class imbalance (see leakage diagnostic before training) - **Impossible travel detection** — geo-velocity feature engineering - **MFA bypass detection** — distinguish FIDO2 anomalies from push fatigue - **Lateral movement detection** — session-graph traversal patterns - **Golden Ticket / Pass-the-Hash** detection benchmarking - **UEBA precision/recall tuning** with calibrated false-positive baselines - **Conditional Access policy effectiveness** simulation - **Zero Trust posture validation** — ZTNA block rate analysis ## Loading the Data ```python import pandas as pd sessions = pd.read_csv("login_sessions.csv") users = pd.read_csv("user_risk_summary.csv") events = pd.read_csv("auth_events.csv") domains = pd.read_csv("identity_topology.csv") # Join session data with user-level risk labels enriched = sessions.merge(users, on="user_id", how="left", suffixes=("", "_user")) # Threat-actor tier classification target (4-class) — see leakage diagnostic y_tier = sessions["threat_actor_capability_tier"] # Binary account-takeover detection target y_ato = users["account_takeover_flag"] # Binary impossible-travel target y_it = sessions["impossible_travel_flag"] ``` For a worked end-to-end example with user-risk-tier classification, stratified splitting, and feature engineering, see the inference notebook in the [baseline classifier repo](https://huggingface.co/xpertsystems/cyb006-baseline-classifier/blob/main/inference_example.ipynb). ## License This **sample** is released under **CC-BY-NC-4.0** (free for non-commercial research and evaluation). The **full production dataset** is licensed commercially — contact XpertSystems.ai for licensing terms. ## Full Product The full CYB006 dataset includes **~1.1 million rows** across all four files, with 12 calibrated benchmark validation tests drawn from authoritative identity security and threat intelligence sources. 📧 **pradeep@xpertsystems.ai** 🌐 **https://xpertsystems.ai** ## Citation ```bibtex @dataset{xpertsystems_cyb006_sample_2026, title = {CYB006: Synthetic Login Activity Dataset (Sample)}, author = {XpertSystems.ai}, year = {2026}, url = {https://huggingface.co/datasets/xpertsystems/cyb006-sample} } ``` ## Generation Details - Generator version : 1.0.0 - Random seed : 42 - Generated : 2026-05-16 14:13:20 UTC - Session model : 6-phase login lifecycle state machine - Benchmark tests : 12/12 passing